Ice-Volcano Interaction at the Subglacial Grímsvötn Volcano,Iceland

In: S.C. Colbeck (ed.): Glaciers, Ice Sheets and Volcanoes: A Symposium Honoring Mark Meier. In Press.
Magnús T. Gudmundsson
Science Institute,
University of Iceland,
Dunhaga 5, 107 Reykjavík.
Abstract

For at least two centuries, volcanic activity in Grímsvötn has been characterized by frequent small eruptions within the composite Grímsvötn caldera and larger, less frequent, fissure eruptions outside the caldera. The caldera eruptions take place within a subglacial lake and rapidly melt the ice above the vents, forming openings in the ice shelf covering the lake. Mounds of hyaloclastites are piled up at the vents, attaining elevations similar to the lake level. Volume of ice melted during these eruptions is less than 0.1 km³. In contrast, the fissure eruption in 1938, which occurred to the north of the Grímsvötn caldera, melted 2 km³ of ice over several days as a subglacial hyaloclastite ridge with a volume of 0.3-0.5 km³ was formed. Simultaneously, meltwater was drained away in a jökulhlaup. In eruptions that break through the ice cover, it appears that the water level at the eruption sites controls the elevation of ridges and mounds formed. For eruptions that penetrate the ice cover outside the caldera, this water level seems to lie several hundred meters below the ice surface prior to eruption. Locally enhanced melting of ice at eruption sites suggest that thermal effects of individual eruptions last 5-20 years.

Introduction

Over the last several hundred years, many eruptions have occurred within ice caps covering parts of the volcanic zones in Iceland (e.g., Thorarinsson, 1974; Björnsson and Einarsson, 1990). Outside the areas presently covered with ice, hyaloclastite ridges and table mountains formed under ice during the Pleistocene glaciation are common landforms. Ice-volcano interaction is therefore an integral part of volcanism in Iceland and jökulhlaups and fallouts of tephra from eruptions within ice caps pose hazards to some areas. In this paper the focus will be on eruptions at the subglacial Grímsvötn volcano. Various data are used to define the characteristics of eruptions. These data include the bedrock topography at eruption sites, data on temporal changes in ice cover and observations of the most recent eruptions. Parts of the work reported have appeared earlier in Gudmundsson and Björnsson (1991).

The Grímsvötn volcano

The Vatnajökull ice cap (8300 km²) covers the central part of the Eastern Volcanic Zone, including several active volcanoes. The most active of these is Grímsvötn in the central part of Vatnajökull. Geothermal and volcanic activity at Grímsvötn has greatly modified the ice surface topography, creating a 10-12 km wide and 200-300 m deep depression. The Grímsvötn depression is a separate ice drainage basin with an area of 160 km² (Björnsson, 1988). Within the depression a subglacial caldera lake is located, covered by an ice shelf that is 250 m thick. Ice flows into the Grímsvötn lake where it is melted at the base of the ice shelf. This results in a rising lake level, eventually leading to a jökulhlaup that partly drains the lake (Thorarinsson, 1953; Björnsson, 1974). Jökulhlaups occur on average once every five years and the volume drained in each event has varied from 0.6 to 4.7 km³ (Gudmundsson et al. 1995). The whole of the Grímsvötn basin is at an elevation well above that of the present equilibrium lines on Vatnajökull. Therefore, basal melting by geothermal heat controls its mass balance, surface ablation plays only a minor role (Björnsson, 1974; Björnsson and Gudmundsson, 1993).

Calorimetric estimates based on water accumulation in Grímsvötn yield a geothermal power of several thousand megawatts (Björnsson et al. 1982; Björnsson, 1983; Björnsson and Gudmundsson, 1993). Fluctuations in geothermal power are closely related to fluctuations in volcanic activity. These fluctuations control the mass balance of the Grímsvötn basin and a reduction in volcanic activity in the latter half of the 20th century has resulted in the buildup of ice in the area and a reduction in the volume of jökulhlaups (Gudmundsson et al., 1995).

The ice cover in the Grímsvötn area is 300-700 m thick but rocks outcrop at the southern rim of the composite Grímsvötn caldera. Exposed rocks in the caldera fault are basaltic tuffs with dikes and sills in places. The subglacial topographic expression of the Grímsvötn volcano is a complex of ridges and mounds, covering an area 10-15 km in diameter and rising 300-800 m above the surrounding bedrock (Björnsson, 1988). The subglacial lake is located within the caldera, which has been in recent times been the center of geothermal and volcanic activity.

Eruptions

Over the last four hundred years activity has fluctuated but eruptions have, on average, occurred once every 10-20 years (Thorarinsson, 1974; Björnsson and Einarsson, 1990). Since 1938, only one minor eruption has been observed, making the latter part of the 20th century the most quiet period since 1600 A.D. Despite the frequent eruptions, direct observations of eruption sites only exist for the eruptions of 1934, 1938 and 1983 (Áskelsson, 1936; Nielsen, 1937; Thorarinsson, 1974; Grönvold and Jóhannesson, 1984). Two of these were small eruptions within the caldera (1934 and 1983) while the eruption in 1938 was a large, almost entirely subglacial, fissure eruption to the north of the caldera.

Caldera eruptions

The craters of the eruptions in 1934 and 1983 were located underneath the southern rim of the caldera (Figs. 1 and 2). In 1934 three craters were active, but two of these were small with minor activity compared to the main crater.

The beginning of the eruption in 1934 was marked with an earthquake swarm. The bedrock at the eruption site at the start of the eruption was at least 100 m below the lake level but the thickness of the ice shelf was 50-80 m. Apparently, the eruption melted its way through the ice shelf in less than half an hour, the time between the first recorded earthquake (of magnitude 4½ on the Richter scale) and the first sighting of an eruption column (Tryggvason, 1960). The eruption column was seen from various parts of Iceland for 8 days and fallout of tephra was detected 100-150 km to the north and the east of Grímsvötn (Thorarinsson, 1974). Some activity was still taking place 15 days after the start of the eruption (Áskelsson, 1936) but two weeks later the eruption had ceased (Nielsen, 1937). At the main eruption site the opening formed in the ice shelf was elongated along the caldera wall, about 1000 m long and 500 m wide. Smaller openings were at other two craters. The total volume of ice melted during the eruption was 40-50 m³. Within the large opening a pile of hyaloclastites had formed (volume 15-20 m³), considered to be the rim of the crater (Fig. 2). Data on the eruption is summarized in Table 1.

The eruption in 1983 was also preceded by an earthquake swarm and 8 hours elapsed from the seismically estimated start of the eruption until an eruption column was seen (Einarsson and Brandsdóttir, 1984). A relatively long subglacial phase in 1983 compared to 1934 was probably due to the combined effects of lower effusion rate and thicker ice cover in 1983. The opening formed had a diameter of 500 m and the volume of ice melted was 20-30 m³. A mound of hyaloclastites with a volume of 6-8 m³, was formed within the opening. The eruption lasted for 5-6 days and fallout of tephra was minute, covering only a few km² close to the vent (Grönvold and Jóhannesson, 1984).

A minor volcanic event, lasting for about an hour, may have taken place in 1984 according to seismic data (Björnsson and Einarsson, 1990). Locally enhanced melting was not observed in Grímsvötn in relation to this event.

Cooling of the material erupted in 1983 took some years. Heat derived from the mound and underlying dike sustained the opening in the ice shelf for about a year. In 1987, four years after the eruption, ice melting at the eruption site was still high compared to pre-1983 values. By 1991, however, ice flow had closed the opening and signs of elevated heat flow had declined. Observations after the 1934 eruption suggest similar thermal effects.

Neither the 1934 nor the 1983 eruption sparked off a jökulhlaup from Grímsvötn. The eruption sites were located within the lake, and melting of floating ice does not affect the lake level. Comparison of the volume of ice melted during these caldera eruptions (0.02-0.05 km³) with that drained in typical jökulhlaups also suggests that their effect on the size and general behavior of the jökulhlaups is minor. However, for some eruptions the reverse may be true, i.e., the release of overburden pressure when the lake is drained may have triggered eruptions (Thorarinsson, 1953). Most Grímsvötn eruptions of the 19th and early 20th century seem to have been sparked off in this way. The eruption in 1934 falls into this category since it started at the end of a jökulhlaup when the lake level had fallen about 100 m. This interaction of the caldera lake and the presumed shallow magma chamber in the crust is a feedback mechanism: The periodic variations in overburden pressure cause periodic tapping of the magma chamber. The magma brought to the surface and intruded into the shallow crust supplies heat to the hydrothermal system that sustains the caldera lake by melting of ice. This mechanism has not produced eruptions since 1934 but earthquakes near the ends of jökulhlaups in 1954 and 1960 (Tryggvason, 1960) may have been due to intrusive activity of this nature.

The fissure eruption in 1938

In 1938 a large and unusually sudden jökulhlaup started on 23 May, reached a peak on 26 May and then receded slowly in about 10 days. Aerial inspection took place on 28 May, revealing the cause of the jökulhlaup: A depression with a volume of 2 km³ (8 km long, 2 km wide and 100-150 m deep) had formed in the ice surface to the north of the Grímsvötn caldera (Fig. 1). A several hundred meters deep hole was located in the northern end of the depression. Meltwater had drained from underneath the depression, into the Grímsvötn lake that was 30-40 m below its critical jökulhlaup level (Gudmundsson et al. 1995). The subsequent steep rise in lake level, possibly coupled with elevated temperature of the meltwater (Björnsson, 1992) sparked off the jökulhlaup.

The depression is considered to have been formed by a subglacial fissure eruption. It only broke the surface of the ice at one location, near the termination of activity, forming a small tephra layer that stained a part of the depression. The ice thickness in the area is 400-700 m and radio-echo soundings have revealed a 70-200 m high bedrock ridge (volume 0.3-0.5 km³) which coincides with the depression (Figs. 1 and 3; Table 1). A common landform in the now ice free areas of the volcanic zones in Iceland are hyaloclastite ridges formed in fissure eruptions during the Pleistocene glaciation. From analogy, Gudmundsson and Björnsson (1991) concluded that the observed subglacial ridge was formed in the 1938 fissure eruption.

If earthquakes occurred during this eruption their magnitude was less than 3½, the detection limit for the area at the time (Brandsdóttir, 1984). Thus, the beginning of the eruption cannot be timed, except that it must have started before 23 May, the time of the onset of the jökulhlaup. Ice surface and bedrock topography to the north of Grímsvötn is such that accumulation of large volumes of water at the eruption site is unlikely. Meltwater was probably drained quite fast southwards into the Grímsvötn lake. Since basaltic eruptions usually have the highest effusion rate soon after the start (Wadge, 1981), it is to be expected that melting was most vigorous in the first days of the eruption. It is therefore probable that a delay due to the buffering effect of the Grímsvötn lake was probably no greater than a few days.

The total heat content of the basaltic magma required to form a 0.4 km³ ridge is more than sufficient to melt 2 km³ of ice (Gudmundsson and Björnsson, 1991). By assuming that the magma was quenched as glass, the total specific heat of the magma can melt 2.5-3 km³ of ice. The latent heat would then be released gradually during alteration of the glass to palagonite. If all the magma was crystallized as pillow lavas, the latent heat of the magma would have been released as well as the specific heat. If so, the total heat available would have been sufficient to melt about 4 km³.

The depth of the depression in the ice surface decreased rapidly after 1938. In 1946, the northernmost part still had some crevasses and a clear depression is evident on air photos from 1954. Apparently, some subglacial melting was still taking place at the eruption site 16 years after the event. Irregularities in jökulhlaup frequency and volume in the decade following the 1938 eruption may be explained by frequent draining of meltwater from small subglacial vaults above a cooling hyaloclastite ridge (Gudmundsson and Björnsson, 1991).

A few earlier eruptions are known to have occurred to the north of Grímsvötn. The eruptions in 1684/1685 (Thorarinsson, 1974) and 1867 (Björnsson, 1988) may fall into this category. The bedrock in this area has many subglacial ridges, some of which may have been formed in these eruptions (Fig. 3). Common to all the ridges is that the ice thickness above the highest points is at least 300-400 m, which is similar to the ice thickness above the ridge considered formed in 1938. The ash-producing eruptions must melt holes through the ice cover and the water level in these holes probably controls the elevation of the ridges formed. Thus, the elevation of the ridges suggests that the vents dispersing tephra over the area, were located in deep depressions (Fig. 4). This further suggests that during eruptions of this type, meltwater is quickly drained away from the vents and collapse of the overlying ice creates the deep depressions. Were eruptions of this type to continue after breaking through the ice cover, subaerially erupted lavaflows might eventually cap the hyaloclastites. The table mountains of the volcanic zones in Iceland are considered formed in this way during the Pleistocene. No table mountains have been found in studies of the bedrock topography of the Grímsvötn area (Björnsson, 1988), suggesting that eruptions usually terminate before subaerially erupted lavas form.

Conclusions

  1. Individual eruptions, lasting several days or weeks, may quickly melt large volumes of ice, demonstrating the highly efficient heat transfer from magma to ice. These eruptions create local depressions in the ice. The depressions disappear gradually, suggesting that the thermal effects of individual eruptions last for 5-20 years.
  2. A feedback mechanism involving eruptions and jökulhlaups may have existed in Grímsvötn. Apparently a decrease in load by a rapidly lowering lake level during jökulhlaups sparked off eruptions within the caldera. In turn the heat from the erupted material and associated shallow intrusions increased the geothermal power, sustaining the high level of ice melting.
  3. The products of the observed subaqueous and subglacial eruptions in Grímsvötn are ridges and mounds of hyaloclastites. The maximum elevation of these formations is controlled by the ambient water level at the eruption site. Within the caldera this level is that of the lake. Elsewhere, it is the water level within deep depressions in the ice formed by rapid ice melting and efficient draining of meltwater. A large fraction of the material erupted after the pile of hyaloclastites has reached the water level, is carried away as airborne tephra.

Acknowledgements

Helgi Björnsson introduced me to the study of subglacial volcanoes and his advice and support is greatly appreciated. A review by Gudrún Larsen improved the manuscript.

References

Áskelsson, J. 1936. On the last eruptions in Vatnajökull. Soc. Sci. Isl., 18, Reykjavík. 68 pp.

Björnsson, H. 1974. Explanation of jökulhlaups from Grímsvötn, Vatnajökull, Iceland. Jökull, 24, 1-26.

Björnsson, H. 1983. A natural calorimeter at Grímsvötn: indicator of geothermal and volcanic activity. Jökull, 33, 13-18.

Björnsson, H. 1988. Hydrology of ice caps in volcanic regions. Soc. Sci. Isl., 45, Reykjavík. 139 pp.

Björnsson, H. 1992. Jökulhlaups in Iceland: prediction, characteristics and simulation. Ann. Glaciol., 16, 95-106.

Björnsson, H., S. Björnsson and T. Sigurgeirsson. 1982. Penetration of water into hot rock boundaries of magma at Grímsvötn. Nature, 295, 580-581.

Björnsson, H., and P. Einarsson. 1990. Volcanoes beneath Vatnajökull, Iceland: Evidence from radio-echo sounding, earthquakes and jökulhlaups. Jökull, 40, 147-168.

Björnsson, H., F. Pálsson and M.T. Gudmundsson. 1992. Vatnajökull, northwestern part, 1:100,000. Glacier surface and bedrock topography. Reykjavík, University of Iceland and the National Power Company.

Björnsson, H., and M.T. Gudmundsson. 1993. Variations in the thermal output of the subglacial Grímsvötn caldera, Iceland. Geophys. Res. Lett., 20, 2127-2130.

Brandsdóttir, B. 1984. Seismic activity in Vatnajökull 1900-1982 with special reference to Skeiðarárhlaups, Skaftárhlaups and Vatnajökull eruptions. Jökull, 34, 141-150.

Einarsson, P., and B. Brandsdóttir. 1984. Seismic activity preceding and during the 1983 volcanic eruption in Grímsvötn, Iceland. Jökull, 34, 13-23.

Grönvold, K., and H. Jóhannesson. 1984. Eruption in Grímsvötn: Course of events and chemical studies of the tephra. Jökull, 34, 1-11.

Gudmundsson, M.T., and H. Björnsson. 1991. Eruptions in Grímsvötn, Vatnajökull, Iceland, 1934-1991. Jökull, 41, 21-45.

Gudmundsson, M.T., H. Björnsson and F. Pálsson. 1995. Changes in jökulhlaup sizes in Grímsvötn, Vatnajökull, Iceland, 1934-91, deduced from in-situ measurements of subglacial lake volume. J. Glaciology, 41, 263-272.

Nielsen, N. 1937. Vatnajökull. Kampen mellem ild og is [Vatnajökull. The fight between ice and fire]. H. Hagerup. København. 123 pp.

Tryggvason, E. 1960. Earthquakes, jökulhlaups and subglacial eruptions. Jökull, 10, 18-22.

Thorarinsson, S. 1953. Some new aspects of the Grímsvötn problem. J. Glaciology, 4, 267-274.

Thorarinsson, S. 1974. Vötnin Stríð. Saga Skeiðarárhlaupa og Grímsvatnagosa [The swift flowing rivers. The history of Grímsvötn jökulhlaups and eruptions]. Menningarsjóður, Reykjavík. 254 pp.

Wadge, G. 1981. The variation of magma discharge during basaltic eruptions. J. Volc. Geoth. Res., 11, 139-168.

Figure Captions

Fig. 1. A map of the ice surface topography in the Grímsvötn area (after Björnsson et al. 1992) showing the location of the subglacial lake and the eruption sites in 1934, 1938 and 1983. The cross-sections shown on Figures 2 (A-B) and 3 (C-D) are indicated.

Fig. 2. Mounds formed in caldera eruptions. Right: The mounds formed in the eruptions in 1934 and 1983. Left: A schematic cross-section showing the relationship between the ice shelf, lake level and the erupting crater (From Gudmundsson and Björnsson, 1991).

Fig. 3. Subglacial ridges to the north of Grímsvötn, formed in fissure eruptions. The approximate form of 1938 ice depression is shown and a broken line shows the likely form of a depression created in the eruption that piled up the ridge at 17 km. Ice surface and bedrock topography from Björnsson et al. (1992).

Fig. 4. Schematic cross-sections illustrating the different stages in the formation and cooling of a hypothetical subglacial ridge. Eruptions may be entirely subglacial and terminate before reaching stage B. In stage C some localized melting still takes place while in D the rigde has lost all its heat. Stages A and B apparently take several days or weeks while stage C may last decades.

Table 1. Data on three Grímsvötn eruptions

_____________________________________________________________________
          No.       Volume    Duration      Duration     Ice melted 
          vents     erupted   of eruption   of subgla-     during
                                          cial phase      eruption
                    (106 m³)  (days)         (hours)     (106 m³)
_____________________________________________________________________ 
          
1934      3           30-40     15-30           >0.5         40-50
1983      1             10       5-6             8           20-30

1938      8 km fissure 300-500    >7          >6 days        >2000
_____________________________________________________________________
Based on:  Thorarinsson (1974), Tryggvason (1960), Einarsson and 
Brandsdóttir (1984),  Gudmundsson and Björnsson (1991).